2011 · 35 citations · 2 references
EngineeringBiomedical EngineeringCardiologyBiomedical Signal ProcessingElectrophysiologyMobile Health InformaticsCardiac AssessmentCardiac Signal ProcessingCardiovascular AssessmentElectrocardiogram AnalysisElectrocardiographyVital SignsAlgorithmic VerificationCardiac Feature ExtractionComputer-aided Electrocardiogram InterpretationEcg Feature ExtractionEcg Quality AssessmentEcg Signal PreprocessingElectrocardiogram Feature ExtractionSignal Quality Assessment
For the application of acquiring ECGs from mobile telephones by unskilled users it would be beneficial if the mobile device could assess ECG quality and inform the user if the quality was acceptable. Using the PhysioNet/Computing in Cardiology Challenge 2011 dataset we identified several ECG features that were commonly observed in the training set ‘unacceptable’ category for algorithmic development: flat baseline (FB), saturation (SA), baseline drift (BD), low amplitude (LA), high amplitude (HA) and steep slope (SS). For the training set with each feature detection applied separately the following scores were achieved: FB 76.2%, SA 80.9%, BD 61.3%, LA 75.6%, HA 74.1% and SS 77.5%. With all features combined a score of 91.4% was achieved. For the test set the algorithm classified 181 records as unacceptable and 319 records as acceptable and the score was 85.7%.
2